Big Auto One

Spectre 8x8 · Uncrewed Ground Systems

The vehicle keeps
going when the
link doesn't

Big Auto One builds the software that makes a ground vehicle uncrewed — remote driving, supervised autonomy, an independent safety core and redundant ground control. It runs on the Spectre 8x8. One crew, one mesh, as many vehicles as the mission needs.

The Spectre 8x8: eight wheels, a flat load deck, and the mission fit on the deck blurred out
Spectre 8x8 by Jackson Duma · deck fit blurred
Operation
Driven, supervised or both in one mission
Resilience
No single failure takes away control
Configuration
Sensors and payloads fitted to the task
Assurance
Every capability driven end to end in simulation

What we build

Everything except
the vehicle itself

The hull is somebody else's business, and a good one is already built. What decides whether an uncrewed vehicle is useful is the crew's ability to command it from somewhere else — through terrain, weather, distance and interference. We build that, at both ends and across the network between them, so the behavior at the edge of contact is designed rather than discovered.

  • Vehicle software — perception, terrain assessment, route planning, motion control and health management.
  • An independent safety core — separate hardware that owns the motion output and never plans anything.
  • Ground command and control — operator stations, mission authorization, and a backup center that can take the vehicle without touching it.
  • The network — a self-forming mesh with no fixed base station, and software that assumes the topology will change.
  • A simulation environment — the same software, on real terrain, before any of it reaches a hull.
Jackson Duma Defense Group

The platform is Jackson Duma's

Spectre is an 8x8 built by Jackson Duma Defense Group, the platform manufacturer we work with. It has been tested by multiple governments and entities, and it is ready for manufacturing today. We do not make hulls. We make everything that turns one into a vehicle a crew can send somewhere and trust to come back.

  • Hull by Jackson Duma
  • Tested by multiple governments
  • Ready for manufacturing today
  • Software by Big Auto One

Capability

What it does

Six things the system is built to do. Each of them runs end to end today against the real vehicle software, the real safety core and the real ground software in the loop.

Drive it from anywhere on the mesh

An operator drives with live video and a terrain picture built onboard. The vehicle enforces its own envelope while they do it: speed comes down for slope, surface and obstacles, and the vehicle stops rather than drive somewhere it has assessed as untraversable — whatever the stick is asking for.

Hand it a point and let it go

Designate a goal on the map, set a corridor, a duration and the health limits you are willing to accept, and authorize it. The vehicle plans its route, drives it, and replans around whatever it meets. The authorization is the contract: pause, redirect or cancel at any time you are in contact.

Carry on through loss of contact

Losing the link is a condition, not a fault. An authorized mission continues inside the geofence and the time limit the operator set, and the vehicle re-establishes contact and reports. Outside of an authorization, loss of link has one defined, enforced behavior, and it belongs to the safety core rather than to the software that was driving.

Run several vehicles on one mesh

Vehicles are peers on the same network, each with its own mission and its own authorization. One crew can dispatch a group to a single objective with a separation between them, or work each vehicle independently. Adding a vehicle adds a node, not a new ground installation.

Degrade instead of failing

Lose a compute unit, a radio, a network switch, a sensor or a command center and the vehicle keeps control and keeps going, with less capability and a slower ceiling. Two independent failures bring it to a controlled stop. The degradation is a designed ladder, not an accident of what happened to still be running.

Hold a limit the planner cannot cross

Geofence, speed ceiling, stop-on-loss-of-link and the arbitration between control sources are enforced on the safety core, independently of the software that plans routes. A planning fault is contained by something that does not plan.

The operator console: a street ahead, the planned route drawn on it, vehicle state across the top and the map, comms and controls down the right
The operator console. What the crew actually works from: the view ahead, the route the vehicle intends to drive, the ground it has assessed, and every control it will accept — including the one that stops it. Regions blurred: compute supplier, link and radio detail

Flexibility

Fitted to the mission,
not to a parts list

The stack is written against interfaces, not against parts. What a sensor is matters far less than what it is required to deliver, and that boundary is where the system is designed to be changed.

Sensing is a contract

The autonomy asks for a terrain model and classified obstacles at a rate and a range. How that is produced — and by what — is a fitting decision, not an architectural one. Change the sensing and the planner does not change with it.

Payload-agnostic

Mission equipment rides on documented power, data and command interfaces, and the vehicle will measure and report what it is actually carrying. A payload change is a configuration item with limits attached, not a software program.

Radio-agnostic

Every radio is a peer and no link is privileged. The gateway will carry a session across whatever paths exist — mesh, relay, a long-haul backhaul — and tell the operator honestly what it has left.

Fits the picture you already run

The vehicle publishes into, and takes tasking from, the common operating picture a unit already uses. A hazard somebody else reported becomes a keep-out the route bends around, without an operator retyping anything.

One stack, many hosts

The same software runs on the vehicle, on a bench with the hardware emulated, and entirely in simulation. Standard robotics middleware where an integrator wants it, and a lean path where they do not.

Limits live in configuration

Every threshold that shapes behavior is a versioned, signed configuration value rather than a number buried in logic. Re-roling a vehicle is a reviewed change to a file with a signature on it.

Where it works

Same vehicle,
different day

The missions below are not brochure scenarios. Each one is a rehearsal we run on real terrain data, with the real software, because the ground decides the route and we would rather find that out before the day.

Sustainment

Moving what people should not have to

Load, route and resupply over ground a crewed vehicle should not be exposed on, with the crew somewhere else.

Overwatch

Watching a route, a site or each other

A static or roving vehicle holding a picture of ground it can see, cross-checked against what another vehicle believes.

Disaster response

Going in when the ground has changed

Flooded streets, burnt canyon roads, a route nobody can plan in advance because the map is out of date by hours.

Urban

Streets, and everything in them

Built ground, moving people and vehicles, and positioning that cannot lean on satellites between the buildings.

Talk to us

Bring us the mission

The useful conversation starts with the ground you need covered, the picture you already run, and what has to keep working when the link goes. We will tell you what that means for the vehicle.